Multi-camera vehicle tracking (MCVT) system is a key technology to build an intelligent city and intelligent transportation system. The MCVT system utilizes roadside monitoring devices and computing platforms to achieve city-level road traffic perception at a finer scale by tracking each vehicle on the road to obtain the trajectory. The existing MCVT system mainly focuses on dealing with the algorithm modules separately, without studying the real-time performance and system architecture of the entire pipeline. However, in real-world applications, many of the tricks in improving evaluation metrics are not applicable since there is endless traffic flow running all the time. Therefore, this paper proposes a real-time framework of the MCVT system to streamline all the algorithm modules with proper data flow. Firstly, a multi-threaded multi-vehicle tracking (MVT) module is developed to detect vehicles in each camera. Re-identification is implemented during MVT to provide the feature for tracking in the same camera and across different cameras. Then, after the vehicle leaves the camera field of view, the feature of vehicle trajectory is calculated by averaging the feature at each frame. Meanwhile, a background thread runs continuously to match the incoming feature of vehicle trajectory with past features by conducting a topology searching method. In the end, the vehicle trajectories with well-matched features are considered to be the same vehicle. The entire pipeline has been validated on two different datasets, the precision is 0.64 in the CityFlow dataset which uses 6 cameras simultaneously. In the customized dataset, the same method can still reach real-time when using 16 cameras at the same time with smaller network modules.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    A Real-time Framework of Multi-Camera Vehicle Tracking System*


    Beteiligte:
    Zhuang, Hanyang (Autor:in) / Chen, Xingyuan (Autor:in) / Lan, Jianglin (Autor:in) / Wang, Ying (Autor:in) / Lu, Zhaoxuan (Autor:in) / Wang, Chunxiang (Autor:in) / Yang, Ming (Autor:in)


    Erscheinungsdatum :

    24.06.2024


    Format / Umfang :

    1816866 byte




    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch



    Multi-camera fusion real-time vehicle tracking method and device

    SONG JIANBIN / JIANG DEHONG / WU YINGJIE et al. | Europäisches Patentamt | 2023

    Freier Zugriff

    Real-Time Multi-Vehicle Multi-Camera Tracking with Graph-Based Tracklet Features

    Nguyen, Tuan T. / Nguyen, Hoang H. / Sartipi, Mina et al. | Transportation Research Record | 2023


    Real-time wide area multi-camera stereo tracking

    Tao Zhao, / Aggarwal, M. / Kumar, R. et al. | IEEE | 2005



    Real-time object tracking system using PTZ camera

    Niu, Hongchuang / Liu, Jia / Yu, Zhengning et al. | IEEE | 2022